On 10/28/14 at 07:02pm, Drew DeVault wrote:
On 10/28/2014 06:55 PM, Dave Reisner wrote:
Does this work with URLs like:
git://example.com/repo.git?tag=v1.2.3
Those don't work with makepkg as it stands now. It doesn't work after this patch is applied, either. If we wanted to make that work, we could do something more sophisticated to take advantage of the --branch option in git.
Not following -- this is even documented to work in PKGBUILD(5). Could you provide an example where this fails using makepkg from git?
Ah, I went and read PKGBUILD(5) and the syntax is `#tag=...`, rather than `?tag=...`. This patch does indeed break that feature. The problem could be avoided by doing the initial clone with --branch, but that fundamentally changes the relationship of the initial clone and the working directory. Currently, the initial clone creates a mirror which the working directory draws from. If we made the mirror use --branch instead, then the working directory would be less of a clone and more of a copy. I don't know the intentions behind this design originally, would that be a bad idea?
My gut says to just drop --depth from this patch and change it to only deal with submodules. The goal of this was to deal with submodules, the shallow change was just a nice-to-have. I could be convinced otherwise, though. Thoughts?
Checking out submodules was previously discussed here: https://lists.archlinux.org/pipermail/pacman-dev/2013-March/016781.html apg